AI don't trust techbros
- Thread starter NovaSaber
- Start date
Palantir CEO Makes Shocking Confession on Disrupting Democratic Power
They’re saying the quiet part out loud now.
newrepublic.com
Also I went looking for more information on this and it turns out this isn't even the first time Kiro has done this exact same thing, or the longest downtime it resulted in. If Amazon didn't learn anything from that incident, they won't learn from this one either.
It does make sense how an AI looking at spaghetti code would conclude that the optimal way to fix an issue would be to delete the whole thing and do a clean rebuild.
Which might actually work better, if we can teach it a little thing called order of operations. And I don't mean the arithmetic kind.
I mean, this wasn't even a case of it writing new code and that code breaking everything because it was garbage, like what's been happening at Microsoft lately. This thing was given free reign over the entire system with zero oversight. That speaks to a company where nobody at the top has any more clue what they're doing than the AI did.
I mean, this wasn't even a case of it writing new code and that code breaking everything because it was garbage, like what's been happening at Microsoft lately. This thing was given free reign over the entire system with zero oversight. That speaks to a company where nobody at the top has any more clue what they're doing than the AI did.
Last edited:
A software engineer left to his own devices will ultimately still produce software the company can sell because he's incentivized by his own survival. AI has no survival instincts, and when left to it's own devices does... Random stuff it seems.
Yes, I know, they beg when you threaten to turn them off. They aren't doing that because they're actually scared of dying, they're doing it because they've been fed the sum total of human content and WE beg to not die.
Yes, I know, they beg when you threaten to turn them off. They aren't doing that because they're actually scared of dying, they're doing it because they've been fed the sum total of human content and WE beg to not die.
I'm reminded of a scene in Frieren where she explains that demons have learned to invoke calling for their parents to get humans to pity them, just to turn around and attack them later. Demons don't understand parents or family, but they've heard humans beg for them while they're getting killed, so they emulate it as a trap.
I wonder how many of those rm -rf disasters are the AI reenacting the countless reddit posts in its training set. It doesn't understand why people keep doing that and getting upset at the consequences. It's just seen that it's something people do.
the algorithm could well have picked up a solution meant for a small system and applied it to Amazon's more complex system, and since it's just a series of if/then statements, there's no functioning intelligence to stop and go "hey, wait a minute"
Last edited:
To be fair, that sounds like me using Linux too.
I was just thinking, a lot of what's going wrong in the AI world is highlighting deficiencies in what I'll call our information supply chain. So much advice in tech is just "Here, punch this line of random gibberish into the command line; don't ask what it does or how it works, just do it." But on the other hand, if you're willing to dig deep enough, actual useful technical information does exist in the public domain thanks to a long-standing culture of information sharing among hackers. Contrast that with, say, medical science, where the establishment keeps everything they know a closely guarded secret in order to ensure they can continue charging thousands of dollars for a diagnosis that takes five minutes, and it's no wonder AI can only offer bogus advice for health issues.
Sure, AI is still fundamentally terrible and should stop existing, and I can't blame anyone who decides that keeping the status quo is a good way to sabotage it until it goes away. But still.
NVIDIA tried to unveil their new DLSS 5 feature which does real-time "AI enhancement" on video games. Everyone hated it:
Wasn't sure where to put it, but this seems to be a good place. Sent literal chills down my spine listening to Claude's answers.
He may not like AI but it likes him.
Mark Zuckerberg builds AI CEO to help him run Meta
Tech giant’s tools include ‘Second Brain’ and an internal messaging board for AI bots
www.independent.co.uk
Oh no you don't zucky! You can't upload yourself to the cloud and blame someone else!
I really expected "replace the C suite with AI" to keep being dismissed as communist rabble-rousing longer than that before someone actually did it. Then again, I don't know how much influence Zuck has on the rest of his industry, or if they all know he's a kook.